Q&A for How to Build a Supercomputer

It is certainly possible to link a number of ordinary PCs together to make a small cluster. However, I do not think that you would be able to do this easily using the existing Windows operating systems as I don't think anyone has really pursued this (probably due to reasons of cost and proprietary issues). You could, however, easily use the same PCs, without affecting the existing O/S either by installing Linux as a "dual boot" system or installing it on multiple USB memory sticks and booting the PCs from them. Linux is really the only choice for doing this as it is the system on which most work has been done in this field, with some versions being specifically designed for clusters.

Yes, but don't expect very good performance. Raspberry Pi's aren't very powerful, but they are silent and output almost no heat, which is ideal.

It really depends on how the power distribution is set up. Too much power can heat it up much more. Not having a good cooling method can also cause heat. Make sure you have a good power distribution method, even if you have to pay more -- it is better than short circuiting your nodes.

If you want to use your laptop as the head node of your supercomputer, just connect it to your ethernet switch via a cable (Wi-Fi is much too slow). Make sure you use a Gigabit (T1000) switch and use all cat6 cables. The laptop must, of course, be running the same Linux operating system as the rest of the cluster, and any laptop is likely to be rather underpowered compared to other processors that might be used. Remember, only software specifically written for distributed computers (parallel computing) will see any benefit from such a system.

It is better that all nodes have the same CPU processing capability. In that way you can easily compute the processing time and also the processing power of the cluster. It also reduces the amount of resources that waste. For example, if there are different CPUs in two nodes, one node may be faster than another. The result is that, in the heavy computing jobs, the faster CPU must wait for the slower one, and this costs extra resources and time.

It is definitely viable to use Ubuntu. It all depends on personal preference. Theoretically, you could use any Linux distro you want.

You need a piece of paper unless you can keep all the numbers in your head. Gather the nodes or pre-build servers and look at their power consumption. If you followed the steps correctly, you should have the same nodes or pre-build servers, so the power output is the same. Now, count how many nodes or pre-build servers you have and then multiply the power output by how many nodes you have. Finally, add 9.

Any software that will work on each individual node.The best software that is free is linux.

Pretty much everything, just faster and more "accurate" than a typical PC. The ones you hear about are for extremely complex math, networking, and/or hosting.

The price can reach a number nobody would like to see, especially for a high end, scientific engineering computer, but it differs depending on what it will be used for.

You could try finding one on eBay.

The VMS would be limited by your own resources, minus some accounting for emulation, but you could try.

Depending on how it's built, you can run a game on one node of a supercomputer. And considering they generally have neither GPUs nor monitors, you will be limited to whatever runs over SSH on the console.

On an HPC cluster, most of the software is looking to run as fast as possible with little to no latency (high-speed low latency). When you run the OS on "bare-metal" or directly on the hardware, there is a direct connection between the application, the OS, and the hardware. When you try to do the same on a virtual system (VMWare, Citrix, Hypervisor, etc), you add a layer of very complex abstraction between the node OS and the hardware. What you will find typically is additional latency and, for a number of applications, compute errors due to that latency. If you are testing your sysadmin skills, by all means, give this a shot. Otherwise, look at lots of hardware.

Theoretically, you could, but connecting to WiFi would be impractical, as it would slow everything down, and consume a lot of the data.

Yes, it will be slower, but since Raspberry Pis emit next to no heat, a normal fan is more than sufficient to keep the hardware cool.

No, not really. FPS is usually GPU bottle-necked, so just get a faster graphics card instead.

You can try, but it may not be successful. You would need a lot of resources, and a lot of training to be able to assemble such a computer.

Virtually anything computing-related. You can host, do mathematics, and calculate Pi for no reason to the trillionth number. Supercomputers are pretty much what the name is; computers, but better.

Yes, if you want to focus your supercomputer on GPU compute. However, cryptocurrency mining takes a toll on GPUs, so keep that in mind.